Emotion Recognition Using Meta-Learning based on Facial Expressions

Emotion recognition is essential in many real-life problems and applications such as computer-human interaction, health monitoring, etc. In this project, we propose a meta-learning approach for developing a computer visionbased emotion recognition system. The aim of the work is to find human emotions from visual inputs and sort them accordingly. Meta learning techniques are used to allow the model to quickly adapt itself to emotion patterns using little training data that is short and compact. Therefore, meta learning technique is important since it allows the algorithm to perform well among different datasets with diverse data sets including unseen emotions (unknown emotions). Emotion recognition should address generalization problems while having ability to handle adaptation issues. It therefore seeks transparency in emotion recognition in addition to overcoming challenges of adaptability. Its objective is to offer answers to practical cases where emotion detection can be applied within dynamic environments. Thus at the end of this project we expect a reliable, faster, more dynamic way of classifying and recognizing human emotions based on visual inputs which highly valuable in understanding human emotional state for multiple other applications.